Pear's latest demo day offered a compact read on where early-stage venture capital is leaning in frontier AI: toward companies that do not merely wrap existing models in a new interface, but instead build the underlying systems that make AI more capable, more efficient, and more deployable in the physical world. TechCrunch, which attended the event, identified five startups that generated the most investor buzz, underscoring a broader shift in the market from generic AI applications to infrastructure, hardware, and model architectures with clearer technical moats.
Spatial AI Bets
Among the companies that caught attention were startups working on spatial models, a category that has become increasingly attractive as investors look for AI that can understand and operate in three dimensions rather than only through text or images. Spatial intelligence is emerging as a key frontier for robotics, industrial automation, simulation, and augmented reality, all areas where standard large language models are not enough. The appeal for VCs is straightforward: if a startup can build a system that accurately maps and reasons about physical environments, it can potentially unlock applications with higher switching costs and stronger defensibility than consumer-facing AI tools.
That interest reflects a larger industry pattern. As foundational model providers continue to improve general-purpose capabilities, investors are searching for startups that can own a specific layer of the stack. Spatial AI sits at that intersection of software and the physical world, where technical complexity can translate into durable value. For early-stage firms, that is especially important in a market where many AI products can be copied quickly and where distribution alone is rarely enough to sustain a venture-scale business.
Chips For Local AI
Another major theme from the demo day was hardware designed to run AI locally. Startups building chips or specialized compute for on-device inference are drawing attention because they address one of the most persistent bottlenecks in AI deployment: cost, latency, privacy, and dependence on cloud infrastructure. Local AI has become more compelling as enterprises and consumers alike seek systems that can process data closer to where it is generated, reducing network delays and limiting exposure of sensitive information.
The investor interest in local AI chips also reflects a practical concern about the economics of the current AI boom. Training frontier models remains expensive, but inference at scale can be even more consequential for companies trying to serve millions of users or embed AI into devices, vehicles, and industrial systems. Startups that can materially reduce compute requirements, power consumption, or latency may find a receptive market among both hardware buyers and strategic investors. At demo day, that proposition appeared to resonate strongly.
Why VCs Are Leaning In
Pear's event offered a reminder that the most compelling AI startups are increasingly those that solve hard technical problems rather than simply package existing capabilities. In the current funding environment, investors are more selective than they were during the early generative AI surge, when nearly any product with an AI label could attract attention. Now, the bar is higher. Venture firms want evidence of proprietary technology, a credible path to deployment, and a market large enough to justify the risk.
That helps explain why the startups that stood out at the demo day were not necessarily the loudest or most consumer-facing. Instead, they were the ones addressing foundational constraints: how AI understands space, how it runs efficiently on local devices, and how it can be integrated into systems that need reliability rather than novelty. Those are the kinds of problems that can support long development cycles and, if solved well, create strong strategic value.
The event also illustrates how demo days remain a useful window into the venture market's evolving priorities. While many AI startups continue to chase software productivity use cases, the most sophisticated capital is increasingly flowing toward infrastructure and enabling technologies. That does not mean application-layer companies are out of favor, but it does suggest that the next wave of breakout startups may come from the less glamorous parts of the stack.
For Pear, the latest demo day appears to have reinforced its position as a backer of technically ambitious founders. For investors, the message was equally clear: frontier AI is no longer just about model size or chatbot polish. The most interesting opportunities may lie in the systems that make AI more spatially aware, more efficient, and more useful outside the cloud.